Cell stiffness predicts cancer cell sensitivity to ultrasound as a selective superficial cancer therapy

Abstract We hypothesize that the biomechanical properties of cells can predict their viability, with Young's modulus representing the former and cell sensitivity to ultrasound representing the latter. Using atomic force microscopy, we show that the Young's modulus stiffness measure is significantly lower for superficial cancer cells (squamous cell carcinomas and melanoma) compared with noncancerous keratinocyte cells. In vitro findings reveal a significant difference between cancerous and noncancerous cell viability at the four ultrasound energy levels evaluated, with different cell lines exhibiting different sensitivities to the same ultrasound intensity. Young's modulus correlates with cell viability (R 2 = 0.93), indicating that this single biomechanical property can predict cell sensitivity to ultrasound treatment. In mice, repeated ultrasound treatment inhibits tumor growth without damaging healthy skin tissue. Histopathological tumor analysis indicates ultrasound‐induced focal necrosis at the treatment site. Our findings provide a strong rationale for developing ultrasound as a noninvasive selective treatment for superficial cancers.

microenvironment, as reflected by intrinsic changes in cell and tissue structure, mechanics, and the biophysical properties of the extracellular matrix. 2,3 For example, malignant cells are easier to deform compared with their noncancerous counterparts because their fewer, less organized F-actin filaments produce a weaker cytoskeletal structure. [4][5][6] Since malignant cells are more deformable, they may possess the ability to migrate through surrounding tissues more easily. 6,7 Squamous cell carcinoma of the head and neck (HNSCC) [8][9][10][11][12] is an aggressive cancer, with patients reporting high levels of disease and treatment-related symptoms affecting basic daily functions, such as speech, chewing and swallowing, and facial expressions. Currently, the most widespread and efficient treatment for superficial cancers is excisional surgery, 13 although local irradiation and topical creams are also applicative. 14 Patient recovery from these ablative procedures may be accompanied by lengthy wound healing processes and esthetic impairment. These unfavorable outcomes highlight the clinical need to develop an efficient treatment that does not harm normal cells and normal tissue function.
The use of ultrasound as a tool in cancer therapy has been studied since the 1940s 15,16 Studies from our laboratory 17 and by others [18][19][20][21] have found that certain malignant cells are highly sensitive to ultrasonic irradiation. Ultrasound produces a variety of nonthermal mechanical bio-effects, 22,23 inducing shear stress 24,25 in cells and stretch/compression distributions in the vicinity of the cellular surface through microstreaming around bubbles, cavitation, and acoustic streaming. 22 Ultrasound pulsing reversibly perturbs the physical and subcellular structures of living cells. 26 Consequently, transient membrane permeabilization (sonoporation) or cell death, depending on the ultrasound conditions, 22,23,27,28 can occur. Moreover, in vivo results from Azagury et al. 17 indicate that the direct application of low-intensity ultrasound to sarcoma tumor reduced tumor growth and increased tumor lysis in mice. However, more widespread and detailed studies are required before low-intensity ultrasound can be used in clinical applications. For superficial HNSCC, such as on the lips and nose, [29][30][31][32][33] the applicability of ultrasound as a treatment modality is expected to be relatively simple, because the ultrasound would be applied topically, as such tumors on a superficial organ can be easily accessed.
Although considerable research has evaluated the role of ultrasound in cancer therapy, historical review 34 and recent comprehensive review 35 showed no effect of ultrasound at all. It is difficult to compare and draw any conclusions from the contradictive results of various investigators, since so many different ultrasound application protocols and tumor model systems have been used. These studies led to our research hypothesis that a single parameter, representing the biomechanical properties of different cell types, can predict their sensitivity to ultrasound treatment. To test this hypothesis, we used atomic force microscopy (AFM) to undertake indentation measurements on different types of superficial cancer cells and thereby examine their deformability, as represented by their Young's modulus, which is a measure of the stiffness of an elastic material. 36 The question of whether the biomechanical characteristics of malignant cells are broadly similar across all tumor types remains unanswered. Consequently, we focused on superficial cancers, particularly on HNSCC, as a model to test our hypothesis.  (Figure 1(a,b), respectively), with higher Young's modulus values indicating stiffer cells. As can be seen in Figure 1(c), the average Young's modulus of noncancerous HaCaT cells is 34 ± 3 kPa, which is significantly higher than the values for UM-SCC47 (25 ± 2 kPa; p = 0.0295), Cal33 (6.2 ± 0.6 kPa; p < 0.0001), and A375 (1.6 ± 0.2 kPa; p < 0.0001).
The actin network, formed by actin filaments (F-actin) or stress fibers, significantly contributes to the mechanical stability (elasticity or stiffness) of living cells, 7,37 and modifications to the actin cytoskeleton during the metastatic process correlate with cell malignancy [38][39][40] The arrangement of fluorescently labeled F-actin filaments in HaCaT, Cal33, and A375 cells was visualized by confocal fluorescence microscopy to verify whether the observed differences in their mechanical behaviors reflect differences in their F-actin network structures.

| Correlation between cells' sensitivity to ultrasound and their Young's modulus
Having established the Young's modulus of the different cell types, we investigated whether it can predict cell sensitivity to ultrasound treatment. We exposed HNSCC cells (Cal33) and noncancerous cells (HaCaT) to different ultrasound operating conditions to identify the ultrasound parameters that cause damage to cancerous cells while being tolerated by healthy tissue. Figure 2(a) is a schematic presentation of our experimental setup for measuring cell viability following ultrasound exposure. Four ultrasound energy levels were tested in this experiment: 2.8, 3.3, 5.6, and 6.6 J/cm 2 . These energy levels were achieved using an ultrasound frequency of 20 kHz, intensities of 0.139 or 0.164 W/cm 2 , and exposure times of 20 or 40 s, while operating at a 50% duty cycle. As can be seen in Figure 2  The identification of a non-molecular cellular parameter that differs between cancerous and noncancerous cells-in this case, Young's modulus, which is a biomechanical measure-potentially opens the way to personalized cancer therapy.

| Ultrasound treatment delays tumor progression in vivo
To validate the potential of ultrasound as a treatment for superficial cancers in tumor-bearing mice, we initially conducted a safety study to evaluate the effect of ultrasound on normal, healthy skin. Since we aimed to evaluate the effect of ultrasound in vivo, in which the ultrasound energy needs to permeate the ultrasound gel (coupling agent) placed above the skin surface and the dense tissue rather than an aqueous medium, the energy level applied for the in vivo experiments was two orders of magnitude higher than the energy level used in the in vitro experiments, mainly due to the large attenuation of ultrasound in the ultrasonic gel evident by the gel temperature increase requiring gel replacement every 30 s as described the materials section.
We utilized an ultrasound application protocol that was previously tested in our lab 41 and found safe for the skin of NOD/SCID mice, namely, operation for 3 min at an intensity of 12.3 W/cm 2 and a 50% duty cycle (corresponding to an energy level of about 340 J/cm 2 ). As can be seen in Figure 3 For the efficacy study, the tumorigenic Cal33 cell line was injected under the skin of NOD/SCID mice. When the tumor reached 3-5 mm in diameter, three different ultrasound intensities were applied, 10.5 11.5, or 12.3 W/cm 2 , for 1 min on a 50% duty cycle every other day. Tumor diameter was measured for the calculation of its volume assuming an ellipsoid shape. The tumor mass was measured following its removal (see Figure 3(b) for experimental protocol). On Day 15, the average tumor volume of the experimental group exposed to 12.3 W/cm 2 was significantly (p = 0.0092) lower than that of the control group. Furthermore, there was a significant statistical reduction in tumor mass between all three experimental groups and the untreated groups ( Figure 3(d1)). The smallest average tumor mass was found in group IV (Cal33 mice treated for 1 min every other day at an intensity of 12.3 W/cm 2 and a 50% duty cycle), with one tumor entirely disappearing in this group. It is important to mention that none of the ultrasound treatments caused any visible damage to the exposed skin. Furthermore, reduced fluorescent signal was observed in the ultrasound treated tumors of Cal33-green fluorescent protein (GFP) mice compared with control mice (Figures 3(d2) and 3(d3)). In Group IV, the area of the tumor comprised of cancer cells was reduced (15% ± 7%) compared with untreated control group (60% ± 6%). The tissue that did not express GFP may be either stromal cells or necrotic tumor cells. These results show that the reduction in tumor volume is proportional to the reduction in tumor mass.
To further optimize the ultrasound treatment protocol to achieve the greatest tumor reduction in the shortest time under in vivo conditions, we examined various treatment repetition schedules to obtain the most effective treatment regime that could safely be administered to each tumor. We therefore examined tumor progression on Cal33 (1) Tumor mass measurements, 15 days after the treatment groups were first exposed to ultrasound, using three different intensities for 1 min on a 50% duty cycle. (2) Florescent scanning of Cal33-green fluorescent protein (GFP) histological sections (GFP labeled green, nucleus labeled blue) for (a) the control group; (b) following ultrasound treatment at 12.3 W/cm 2 every other day for 15 days. (3) GFP fluorescent signal analysis (using the ImageJ program) of the control group and a treatment group exposed to an ultrasound intensity of 12.3 W/cm 2 after 15 days of treatment. Statistical significance was calculated using t test, **p < 0.01 mice following ultrasound application at 12.3 W/cm 2 for 1 min on a 50% duty cycle once every other day compared with once a day, and with twice a day treatments. The tumor volume growth kinetics ( Figure 4(a)) indicate that the repetition of ultrasound treatment is associated with enhanced reduction in tumor volume (and consequently with reduced growth). The greatest differences in tumor volume (Figure 4(a)) and mass (Figure 4(b)) were obtained between the control group and the group exposed to ultrasound twice a day (Group VII).
During all the in vivo experiments, no abnormal behavior of the mice was observed throughout the treatment of 14 days. In addition, no effects were seen on the skin or abnormal mortality of the mice.

| Ultrasound treatment-induced necrosis in tumors
To understand the effect of ultrasound treatment (twice a day) on tumor mass and volume, a pathologist evaluated all the tumor cross sections 48 h and 11 days after the treatment groups were first exposed to ultrasound. The visual difference between the control and treatment groups was located in the area of necrosis (AON) (  3 W/cm 2 intensity and a 50% duty cycle) for 11 days on different treatment repetition schedules: ultrasound exposure every other day; once a day; or twice a day. (b) Tumor mass measurements 11 days after the treatment groups were first exposed to ultrasound. Statistical significance was calculated using one-way ANOVA test *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 platform for the personalized, noninvasive therapy of superficial cancers by means of achieving the selective death of cancerous cells.
We used AFM to spatially map biomechanical properties across the surfaces of single cells and thereby obtain the mean Young's modulus values of a population of cells. Cells exhibit internal heterogeneity (for example, the nucleus is about 4-10 times stiffer than the cytoplasm [43][44][45] ) and therefore we chose to perform the measurements using a probe having a spherical geometry to increase the contact area and decrease scattering. The appropriate model that corresponds to the sphere indenter is the modified Hertz model for living cells. 35,46 We found that highly aggressive cancer cells, such as melanoma without the need for external intervention, such as cavitation nuclei or microbubbles. All these natural phenomena, which increase as the ultrasound energy level increases, can trigger biophysical effects, such as microstreaming, microjetting, and free-radical formation, which may affect cell viability. 62 The results show that the ultrasound penetrated the skin safely without causing damage to the healthy tissue and produced selective apoptosis of the cancerous cells only.
We suggest the difference in cancerous versus healthy cells structure and as a result their mechanical properties as presented by their modulus of elasticity, is the main cause for the selective difference in F I G U R E 6 Ultrasound treatment-induced necrosis in tumors: (a) Effect of ultrasound treatment (1 min operation time at 12.3 W/cm 2 intensity and on a 50% duty cycle) on the necrotic area as a percentage of total tumor area (AON%) in groups treated according to different treatment repetition schedules compared with the control group, measured: (1) 48 h and (2) 11 days after first ultrasound application to the treatment groups. (b) (1) Tumor kinetics in the control group and in the treatment groups after 48 h (red) and 11 days (blue) of a twice a day ultrasound treatment schedule; and (2) two-way ANOVA comparing the two treatment durations in terms of the energy level row factor, the cell viability column factor, and the interaction between them, where **p < 0.01, ***p < 0.001, ****p < 0.0001, and ns indicates a nonsignificant result the effect caused by the shear stresses generated by ultrasound.
Resulting in softer cells being more susceptible causing membranal rapture or porosity resulting in apoptosis.
Recent studies have revealed the potential of using ultrasound to activate an immune system response against cancer. [63][64][65] One approach is to deliver immune stimulating agents to tumors by applying ultrasound to ultrasound-sensitive carriers (e.g., tumor antigens or genes), 66 whereas another approach aims to use the mechanical or thermal effects of ultrasound to enhance immune responses. 67 These approaches endeavor to achieve immune modulation. The field of therapeutic immunomodulation is young and the mechanisms whereby ultrasound affects immune response are still not fully understood 57 The current study was performed on mice lacking an immune system, and therefore the results do not reflect any effects of ultrasound exposure on the immune system.
Solid tumors are often first diagnosed by palpation, which may suggest that tumor tissues are more rigid that surrounding healthy tissues. Paradoxically, individual cancer cells are softer than their healthy counterparts. 68 It follows that the correlation between Young's modulus and cell viability following ultrasound application may differ for tissue compared with cells. Nevertheless, our findings indicate that stiffness at the level of the individual cell is the key to selective ultrasound-induced cell death.
Although we found a significant difference in tumor volume between the highest ultrasound intensity treatment group and the untreated control group and, in the treatment group, one of the tumors completely disappeared (Figure 3(d1)), the tumors continued to grow in both groups (Figures 3 and 4). This suggests that the treatment repetition schedules studied were not sufficient to eradicate the tumor and, therefore, additional work is necessary to optimize treatment for complete tumor eradication and the prevention of regrowth.
We showed that ultrasound treatment produced a quantitative effect on superficial tumor progression in vivo. These results, which are consistent with our previous report on breast cancer, 17

| Cell lines and culture conditions
The human keratinocyte cell line (HaCaT) was grown in MEM and supplemented with glucose (4.5 mM), FBS (10% vol/vol), L-glutamine (2 mM; 1% vol/vol), and penicillin-streptomycin (100 μg/ml penicillin and 100 μg/ml streptomycin; 1% vol/vol) in an incubator under a 5% CO 2 atmosphere at 37 C. The cells were split every 2-3 days to prevent overpopulation as follows: the culture medium was removed from the flask and the cells were washed with filtered PBS. Cells were disconnected from the flask after the addition of 2 ml trypsin-EDTA and 10 min in an incubator. Following incubation, growth medium (10 ml) was added. The suspended cells were pipetted three to six times and divided into three flasks (4 ml each). Fresh medium was added to a total volume of 12 ml in each flask. The cells were returned to the incubator for 3 days for further proliferation.

| Effect of ultrasound on cell viability in vitro
Cells were seeded at a density of 160,000 cells/ml in a 12-well plate (each well contained 1 ml of culture medium) and their viability was tested using PB reagent, as described above. Afterward, the plate was washed with filtered PBS, filled with 1 ml of fresh medium and

| Effect of ultrasound on tumor reduction: in vivo efficacy studies
Ultrasound treatment was carried out as previously described by Azagury et al. 17 The current study (IL-80-12-2015) was approved by the Institutional Review Board for animal welfare. Briefly, NOD/SCID mice aged 6 weeks old were injected subcutaneously with 100 μl of 1 Â 10 6 Cal33 HNSCC cell line/100 μl of PBS at two points on their backs. The ultrasound treatments started when tumors reached 3-5 mm in diameter (about 1 week after the injection), as measured manually by a caliper. Tumors that did not reach the appropriate size were not taken in account.
For ultrasound treatment, a cylindrical glass chamber (1.6 cm diameter) was placed over the tumor on the back of each anesthetized mouse and filled with ultrasound gel (3 ml at a temperature of $4 C).
The ultrasound probe was positioned 1 cm from the surface of the skin without touching the chamber walls. The ultrasound (QSONICA, 700 W, 20 kHz) was operated in an intensity range of 10.5-12.3 W/cm 2 for 1-3 min on a 50% duty cycle using a probe with diameter 1.3 cm. Mice were anesthetized by injection of 100 mg/kg ketamine and 10 mg/kg xylazine before application of ultrasound.
Groups that were exposed to the ultrasound more than once per day, requiring a total longer anesthesia per day, were connected to an isoflurane anesthetic system (SomnoSuite, low-flow anesthesia system, from Kent Scientific Corporation) throughout the second sonication procedure. To minimize thermal effects, the ultrasonic gel was replaced with fresh gel every 30 s. During the procedure, before ultrasound application, the gel was kept inside an ice water bowl. After the ultrasound was turned off, the skin was cleaned with Septol.
For the safety experiments, healthy 6-week-old NOD/SCID mice (n = 2) were treated with ultrasound at an intensity of 12.3 W/cm 2 for 3 min on a 50% duty cycle. Immediately after treatment, samples of the exposed skin were taken for histology examination.
To evaluate the effect of ultrasound on tumor reduction, different ultrasound intensities (10.5, 11.5, and 12.3 W/cm 2 ) and treatment repetition rates (every other day, every day, and twice a day), were applied for 1 min on a 50% duty cycle. Cal33 mice (n = 43) were randomized into groups: (I) untreated (control) (n = 9); treatment every other day at (II) 10.5 W/cm 2 (n = 4), (III) 11.5 W/cm 2 (n = 4), or (IV) 12.3 W/cm 2 (n = 10); (V) treatment every day at 12.3 W/cm 2 (n = 7); and (VI) treatment twice a day at 12.3 W/cm 2 (n = 9). During the experiments, tumor width and length (diameters) were measured manually using a caliper. Tumor volume was calculated using the ellipsoid volume equation under the assumption that the depth of the tumor is equal to the smaller diameter value. After 2 days, three mice from groups I, IV, and V, and four mice from group VI, were sacrificed. After 11 days, three mice from groups I and IV, four mice from group V, and five mice from group VI were sacrificed. After 15 days of treatment, three mice from group I, and four mice from groups II, III, and IV were sacrificed.
The tumors were removed and washed with PBS. All the tumors were weighed (except for the tumors that were taken after 2 days for necrosis analysis) and transferred into 4% (wt/vol) PFA in PBS for 1 h/1 mm 3 of tumor volume. Afterward, all the tumors were transferred into 70% ethanol until histology analysis was performed.

| Histology
For histopathological preparation 4% (wt/vol) formalin-fixed paraffinembedded HNSCC tumors were cut to 4 μm sections, mounted on microscope glass slides, and heated overnight at 65 C in a drying oven. Following dehydration, slides were stained with hematoxylin and eosin (H&E), scanned by a Panoramic MIDI II scanner (3D Histech) and analyzed by a pathologist. Necrotic areas within treated tumors were morphologically evaluated. First, the AON was marked and was calculated in arbitrary units using the ImageJ and CaseViewer programs, after which the AON was calculated as a percentage of the entire tumor volume (AON%). Morphological characteristics of necrosis consisted of areas of atypical mitosis, lymphocytes, fibrin, acute inflammation, and tissue loss. Results are presented in AON%. Statistical analysis was carried out by GraphPad Prism 7.03 software, significance set at p = 0.05.

| Statistical analysis
Statistical analysis was performed using GraphPad Prism 7.03 software, presented as mean ± SEM. All cellular experiments were repeated at least three times. For experiments involving less than two groups, oneway ANOVA was used. For experiments involving two groups, a twotailed Student's unpaired t test was performed to compare the control versus treatment groups. For experiments involving more than two groups, two-way ANOVA was used. Values of p ≤ 0.05 were considered significant. For pathological analysis, H&E images were analyzed by Panoramic Viewer Histoquant software (3D Histech), and a one-way ANOVA test was performed to compare control vs. treatment groups. validation; writing-original draft; writing-review and editing.